Most of the bugs and version problems I've seen are in the Java GUI libraries. It can be difficult to write GUI code that works the same way on all of your supported platforms. There isn't much of a way around it other than a lot of testing. Code to a common version (1.3 now?), but test on several versions and platforms.
Allocating memory in Java is expensive, so two techniques that help with performance are reusing objects rather than creating new ones, and using object pools. We improved performance over 25% on a product that parsed genomic data by recoding that way. We used an optimizer to guide our performance tweaks and validate the results. Also starting the JVM is expensive, so when you're connecting processing steps together don't use Java like a scripting language. The JVM spends a lot of time loading and verifying its class libraries when it starts up, which can be overly redundant. Rather than piping results between processes, it's better to have one process. A C application will often execute faster than Java, but Java is faster to develop and the resulting application is more reliable. Best Regards, Brian King _______________________________________________ Biojava-l mailing list - [EMAIL PROTECTED] http://biojava.org/mailman/listinfo/biojava-l
